Anti-Reflective Coating (Arc)

# Anti-Reflective Coating (ARC)

An anti-reflective coating (ARC) in photolithography is a deliberately engineered optical layer used to reduce reflected exposure energy and the interference it creates inside photoresist. A bottom ARC (BARC) lies between resist and the underlying stack; a top ARC (TARC) lies above the resist. Their common purpose is not simply to make a surface “dark,” but to make the resist image less sensitive to film thickness, substrate reflectivity, topography, angle, polarization, and wavelength.

ARC design is a multilayer electromagnetic problem coupled to coating, bake, develop, and etch integration. The relevant material property is the complex refractive index, conventionally written $\tilde n=n-i k$ or $n+i k$ depending on the field convention. The real part controls phase velocity and interface mismatch; the extinction coefficient controls absorption. Thickness determines phase and propagation loss. A useful ARC must deliver the desired optical response while remaining coatable, uniform, defect-controlled, chemically compatible, and transferable through the patterning flow.

## Reflection, interference, and the swing curve

At each interface, part of the incident electric field is transmitted and part is reflected. At normal incidence between simple nonabsorbing media, the amplitude coefficient is

but a production lithography stack requires complex indices, multiple reflections, absorption, oblique incidence, polarization, and partially coherent illumination. Transfer-matrix or electromagnetic solvers propagate the complex field through every layer. The reflected fields add as vectors: their magnitudes and phases determine the total reflectance at the resist boundary.

Without sufficient reflection control, upward- and downward-propagating fields interfere inside the resist. The through-thickness dose can then oscillate, producing standing-wave structure in the chemical image and sidewall. As resist or underlying-film thickness changes, the accumulated phase changes and the effective exposure at the resist moves through maxima and minima. Printed CD therefore oscillates with thickness—the swing curve. Topography and wafer-scale thickness variation convert this optical sensitivity into spatial CD variation.

A single quarter-wave intuition can be useful for a lossless one-layer coating at one angle and wavelength, but it is not a universal BARC recipe. Absorbing films have complex phase thickness; real substrates contain several films; scanner illumination spans angles and polarizations; resist and ARC properties change with bake and wavelength. The correct optimization minimizes a defined output—such as reflectance, absorbed-energy variation, or printed-CD swing—over the qualified distribution of stacks and process conditions.

## BARC and TARC solve different boundaries

BARC attacks reflection from beneath the resist. An absorbing BARC attenuates light on the trip toward and away from the substrate, while its index and thickness rotate interface contributions toward cancellation. Organic spin-on BARCs offer planarization, track compatibility, and potential resist-like removal. Silicon-containing, oxide/nitride-like, amorphous-carbon, or other inorganic layers can combine optical control with hard-mask or etch functions, but they impose deposition and plasma-transfer requirements.

TARC modifies the upper optical boundary. It can reduce the resist/ambient reflection and associated interference. In immersion lithography a topcoat may also control contact with immersion fluid, resist-component leaching, water uptake, and defectivity; those integration functions can dominate its value. A topcoat is not automatically an optical ARC, and an optical TARC is not automatically a qualified immersion barrier. Its dissolution, developer compatibility, refractive index, uniformity, residues, and interaction with resist must all be verified.

BARC and TARC are not interchangeable. A top coating cannot generally extinguish a strong buried-stack reflection by itself, while a BARC does not manage every top-interface or immersion-fluid interaction. Some stacks need one, both, or neither. The answer follows the wavelength, resist, substrate, imaging condition, and process integration—not a node label.

## Material and integration requirements

Optical performance is necessary but insufficient. An ARC material also needs:

• uniform thickness and optical constants across wafer, lot, and shelf life;
• coating or deposition behavior compatible with topography and defect targets;
• no harmful intermixing, dewetting, poisoning, acid/base exchange, or solvent attack;
• thermal stability through soft bake, exposure, PEB, and downstream steps;
• adequate adhesion without creating rework or residue problems;
• a controlled etch-open or removal path with sufficient selectivity;
• minimal added line-edge roughness, footing, undercut, faceting, or CD bias;
• compatibility with the hard mask, target film, clean, and defect-inspection strategy.

Organic spin-on BARC can smooth modest topography but may exhibit thickness and composition changes during bake, solvent interaction with resist, and limited etch resistance. Inorganic or silicon-rich ARC can provide a robust transfer layer but may require dedicated deposition, etch, endpoint, and strip control. A dual-function silicon-containing ARC/hard-mask stack trades fewer optical interfaces against more demanding plasma integration. Material labels alone do not predict performance; measured $n(\lambda)$, $k(\lambda)$, thickness, density, composition, and post-bake state belong in the optical model.

## ARC-open and pattern-transfer consequences

A BARC commonly remains beneath developed resist openings and must be opened before the target film can be etched. This step is not optically visible in the original reflectance optimization, yet it can dominate the transferred result. Lateral attack widens an opening; polymer deposition narrows or foots it; insufficient overetch leaves residue; excessive ion or radical exposure consumes resist and changes the profile. Across-wafer ARC thickness and plasma uniformity become CD-uniformity inputs.

The ARC-open recipe must be designed with the resist and target etch. Key quantities include open rate, resist selectivity, target-film selectivity, endpoint behavior, overetch margin, microloading, aspect-ratio dependence, sidewall chemistry, and removal of residual ARC. If the ARC doubles as a hard mask, its post-open profile and roughness become the mask for all later transfer. After-develop inspection cannot establish that this transfer was successful; paired after-open or after-etch measurements are required.

Rework strategy should be decided before qualification. A resist misprint over a spin-on organic BARC may permit a compatible strip and recoat, but intermixing, ARC damage, or altered surface energy can prevent simple reuse. An inorganic ARC may survive resist strip yet accumulate plasma or clean history. Rework approval needs evidence for optical constants, thickness, adhesion, particles, residues, and final transfer—not merely a visually clean wafer.

## Modeling and experimental qualification

Start with measured layer-specific optical constants at the exposure wavelength and relevant bake state. Build the actual underlying film stack, including thickness distributions and topography. Optimize ARC composition and thickness over the expected variation rather than a nominal stack. Then challenge the model experimentally with thickness wedges or split wafers, exposure/focus matrices, representative pitches and orientations, and reflective-stack corners.

Useful outputs include:

Reflectometry or ellipsometry can monitor film thickness and optical properties, but inverse fits may correlate thickness, $n$, $k$, roughness, and underlying-stack parameters. Reference methods and physically bounded models are needed. CD-SEM, OCD/scatterometry, cross-sectional SEM/TEM, AFM, defect inspection, and residual-film measurements see different aspects of the process. Their measurands and sampling populations should be declared.

A designed experiment should separate ARC thickness, bake, resist thickness, substrate-stack corner, exposure dose/focus, and ARC-open conditions. Randomization protects against time drift; replication measures noise; module and material-lot splits test transferability. The optimum is a robust region, not the single wafer with minimum fitted reflectance.

## Manufacturing control

Production control links ARC lot and age, dispense or deposition module, coat/bake history, thickness and optical-constant maps, substrate-stack genealogy, resist and exposure conditions, develop module, ARC-open chamber, endpoint, and final metrology. Coordinate-preserving maps distinguish radial coating or plate signatures from field/layout effects and plasma-chamber fingerprints.

Feedback should follow causality. A slow, confirmed thickness drift may justify a bounded coating adjustment. A shift in $n/k$ can require material or bake investigation rather than thickness compensation. A final-CD error caused by ARC-open microloading should not be hidden with scanner dose. Faults such as dewetting, particles, dispense interruption, heater failure, optical-constant excursions, endpoint failure, or residue require hold-and-investigate logic.

BARC, TARC, underlayer, hard mask, and topcoat names describe position or function, not guaranteed performance. The disciplined question is: which reflected fields are being controlled, over which stack distribution, and at what integration cost? An art-level ARC solution minimizes sensitivity of the printed image and preserves that advantage through etch, removal, rework, and high-volume manufacturing.